GPU Performance Architect

AMD AMD · Semiconductors · Folsom, CA · Engineering

This role focuses on GPU performance architecture, analyzing and optimizing AI/ML applications and LLM kernels. The candidate will work on simulation models, identify bottlenecks, and propose solutions for performance improvements, with a focus on graphics cores and SoCs. Experience with LLM kernel development using PyTorch, Triton, CUDA, or HIP is preferred.

What you'd actually do

  1. Work on workload/competitive analysis of contemporary and futuristic game/AI/ML applications
  2. Identify complex technical problems, break them down, summarize multiple possible solutions, and help the team make progress
  3. Work with architects to understand bottlenecks in graphics cores and SoCs
  4. Analyze existing and emerging graphics/compute paradigms and algorithms
  5. Implement and run simulation models - i.e., both RTL and high-level simulation to estimate potential gains

Skills

Required

  • Excellent C/C++/Scripting (Python, etc.) experience
  • Knowledge of GPU architecture and Compilers

Nice to have

  • Experience spanning architecture, performance analysis and AI/ML/graphics/compute algorithms
  • Experience in Kernel development of LLMs using PyTorch, Triton, CUDA, HIP etc.
  • Knowledge of Graphics/Compute APIs (DirectX/Vulkan/CUDA/HIP etc.)
  • Experience with AI/ML use cases and demonstrated ability to effectively leverage and collaborate with AI agents in modern engineering workflows
  • Experience in GFX profiling tools is a plus (PIX, RenderDoc, AMD tools, etc.)

What the JD emphasized

  • Kernel development of LLMs using PyTorch, Triton, CUDA, HIP etc.

Other signals

  • GPU Performance Architect
  • AI/ML applications
  • Kernel development of LLMs
  • AI agents in modern engineering workflows